Present: Glasgow: Sam, Gareth IC: Dave, Duncan Cambridge: John, Jeremy Liverpool: Stephen, John RAL: Brian, Jens (chair+mins) Oxford: Ewan Lancaster: Matt Apologies: Edinburgh: Wahid (on leave) 1. xroot "documentation" - where is the documentation... - how to make it publish (probably not essential?) - how "federations" work? xroot also requested for T3 sites. DPM with xroot support. If a file isn't available at a site, then it falls back to a UK redirector, then Europe, then finally world wide. Which CMS sites have this and which still need to publish it? Documentation is mostly there, but may not be visible to people outside CMS. Situation is evolving, and the times they are a-changing. CMS will present performance numbers. ATLAS are also planning to go down this route. More ATLAS sites have DPM than do CMS sites. DPM xroot implementation is fairly new and the YAIM configuration isn't quite in the package so some manual configuration is necessary. How is xroot secured - with certificates? It is used for reading, not for writing. ALICE move their data with xroot, though, so need write access (token based authentication). Without certificates, data may have to be world readable (or by whatever the xroot account uses). ATLAS central have a set of requirements on how secure the data is, and metadata is OK if you know where the file is, but they do mind world readable. This may be true for CMS as well. Would it be sufficient to just know the filename? This could improve performance. So everybody will need to support xroot - ATLAS and CMS together is to a first approximation. StoRM sites - how do they support xroot? Being investigated, it may be sufficient to just run an xrootd server. Should this feed into the requiremtns for T1 "EOS"? It already is, but may be a requirement or desirable. Does xroot have to be published in the BDII? Maybe not. How does CMS interact with the storage group - we discuss this every so often, but should we do better? 2. Events for September - which looks like a Very Busy Month(tm), once again - who is going to what? Like EGI-TF for example. ... someone from GridPP should go? Chris Walker sometimes attends. Dave Colling might go. John Gordon is going, so is Claire. All hands - digital something - no one is going. And everybody's favourite event, GridPP, is in September! Expecting experiment experiences explained expediently? If so, do we need to prepare something, to make the most of the meeting - particularly regarding the... 3. ...ongoing DPM support planning - DC would like to have a roadmap for decisions. ops list may be the best place to circulate it. May also be better to send it to named people. Not that there is anything secret but it may be better to keep the discussion within GridPP at the moment. By the GridPP meeting, we aim to have a roadmap and to know what the problems are. A proposal. We should know what the technical stumbling blocks are, how to make progress with them, what sort of other questions we need to ask (effort from other countries, requiremetns from experiments). We are of course happy to keep doing what we do; we're still resolving the unknowns. Also get as much effort from the experiments. 4. One more attempt at rounding up loose ends - if one can round up loose ends - if there is time - KeyDocs and suchlike. Postponed till next week. 5. AOB [09:59:36] Stephen Jones joined [10:00:12] David Crooks joined [10:01:13] Brian Davies joined [10:01:37] Duncan Rand joined [10:01:43] John Bland joined [10:02:33] Jeremy Coles joined [10:04:41] Ewan Mac Mahon joined [10:12:44] Ewan Mac Mahon Aren't we pretty much going to make everyone's data world readable by WebDAV anyway [10:14:08] Sam Skipsey So, the dCache and DPM WebDAV implementations allow (and enforce) HTTPS security on connections. [10:14:27] Sam Skipsey So, they don't make the data world-readable. [10:15:02] Ewan Mac Mahon But does that do VOMS-y stuff for the VO membership (with gridsite?) at the moment, or is it just 'you have a grid cert'? [10:16:44] Stephen Jones So are we looking at one XROOT baseline that suits both ATLAS and CMS at each GridPP site thatruns DPM? Is that it? [10:18:12] Matt Doidge joined [10:18:37] Ewan Mac Mahon AIUI the basic xrootd service is VO agnostic, then each VO that's going to do full federated redirection needs to have a service on the head node dedicated to them, [10:18:46] Sam Skipsey Yes [10:19:06] Ewan Mac Mahon So at Oxford we have a general xrootd service, plus the head node has an ATLAS specific redirector as well. [10:19:17] Ewan Mac Mahon We could also add a CMS redirector or whatever as well. [10:19:20] Sam Skipsey The redirectors need to point at the next level of the redirection hierarchy to them, which will be a different node for different VOs [10:21:47] Ewan Mac Mahon What I'm not clear about, given the whole 'not in the BDII' thing, is how anyone actually finds out that we've got an atlas xrootd service. [10:24:34] Sam Skipsey Well, if you turn on monitoring, then the upper levels of the hierarchy get reports from your service. [10:25:30] Sam Skipsey (Although, I emailed Michel Jouvin about the "not in the BDII" thing, since it's clear from the source of dpm-listspaces that at the moment it doesn't check for much other than rfio and gsiftp) [10:26:20] Jeremy Coles http://digital-research.oerc.ox.ac.uk/ (the All Hands link?) [10:27:11] Jens Jensen Yep, that's it. [10:27:38] Ewan Mac Mahon Thinking about it, I'm not really sure how my xrootd redirector knows about any other xrootd services either. I can see how the LFC lookup finds the location (SURLs?) of missing files, but it's not obvious what happens next. [10:28:05] Ewan Mac Mahon Particularly when most atlas sites that might have files that we're missing aren't going to be running xrootd. [10:28:31] Ewan Mac Mahon But I'm sure all this will become more apparent once we actually start getting people to use it. [10:28:54] Sam Skipsey Right, the xrootd hierarchy itself knows about xrootd services "below" it. [10:29:15] Sam Skipsey So, the UK redirector knows about the xrootd services that have registered with it. [10:29:52] Ewan Mac Mahon So will ours have done that automatically? [10:30:15] Ewan Mac Mahon And if so, how will it have known where to find the UK redirector. [10:30:28] John Bland left [10:39:54] Sam Skipsey The UK redirector is the Federated redirector you configured it to talk to, Ewan [10:40:14] Gareth Roy left [10:40:15] Sam Skipsey left [10:40:16] David Crooks left [10:40:16] John Hill left [10:40:17] Jeremy Coles left [10:40:17] David Colling left